It's just statistics. I personally think it would be more clear if the SIMD instructions were actually counted as SIMD instructions for x86, but it doesn't really matter. It's not a question of correctness, just a question of how you want to evaluate things.
If it's an easy change, and you want to, posting a patch on reviewboard ( reviews.gem5.org) will let all of the gem5 developers make comments on how we should count x86 SIMD instructions. Cheers, Jason On Fri, Jun 24, 2016 at 1:40 PM Ayaz Akram <aaq...@gmail.com> wrote: > Hi Jason, after looking into x86 SSE instructions in gem5 I have figured > out that most of them are decoded into micro-ops which operate on floating > point registers. Also, I looked at the discussion at the following link: > > http://comments.gmane.org/gmane.comp.emulators.m5.devel/4404 > > After going through these, will I be wrong in concluding that labelling of > SSE micro-ops as floating point micro-ops is correct for most of the cases > ? I'm not sure how close is this implementation to real x86 processors, > though. > > > > On Wed, Jun 22, 2016 at 2:12 PM, Ayaz Akram <aaq...@gmail.com> wrote: > >> Thanks Jason. I'll try to fix it and possibly submit a patch. >> >> >> >> On Wed, Jun 22, 2016 at 9:30 AM, Jason Lowe-Power <ja...@lowepower.com> >> wrote: >> >>> To directly answer your question: The labeling of instruction in gem5 is >>> not accurate at all. Feel free to fix the labeling and submit a patch to >>> reviews.gem5.org, though! We'd welcome the fixes. >>> >>> BTW, you're correct that SIMD instructions are labeled as floating >>> point instructions. I've noticed that too. >>> >>> Cheers, >>> Jason >>> >>> On Tue, Jun 21, 2016 at 3:42 PM Ayaz Akram <aaq...@gmail.com> wrote: >>> >>>> I am running a small C program on gem5. Disassembled binary of the >>>> program shows a lot of SSE instructions, but simulator's output stats.txt >>>> does not show any SIMD operation type committed. On generating the >>>> execution trace from gem5, I can see that those SSE instructions are >>>> decoded into operations and labelled as FloatAdd or other similar >>>> instructions. My question is, how accurate is the labelling of operations >>>> in gem5 for x86 ? I mean it seems a bit misleading to see that no SIMD >>>> instructions are committed , while the benchmark has some of them. >>>> >>>> Thanks >>>> _______________________________________________ >>>> gem5-users mailing list >>>> gem5-users@gem5.org >>>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >>> >>> >>> _______________________________________________ >>> gem5-users mailing list >>> gem5-users@gem5.org >>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >>> >> >> > _______________________________________________ > gem5-users mailing list > gem5-users@gem5.org > http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
_______________________________________________ gem5-users mailing list gem5-users@gem5.org http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
